IDC and Lenovo study finds most AI proof-of-concept projects do not reach production due to organizational and technical readiness gaps

Decision brief

What changed

A joint IDC-Lenovo study reported that 88% of AI proof-of-concept projects do not make it into production, attributing the drop-off to unclear objectives, unprepared data, and limited in-house agentic AI expertise. Coverage also highlighted a “build-vs.-run” gap, where teams can design AI agents but are not ready to operate them reliably inside business processes.

Why it matters

For business leaders, this shifts the decision focus from funding more pilots to testing whether the organization is actually ready to run AI in production with governance, data quality, and operating discipline. The coverage suggests that value realization depends less on model experimentation alone and more on production controls such as monitoring, logging, runbooks, human oversight, and least-privilege access, which affects budgeting, staffing, and deployment standards.

Affected roles

CEO CFO COO CTO CISO

Evidence

The central claim comes from a joint IDC and Lenovo study as reported by The New Stack, which specified the 88% figure and named the main causes and recommended production practices. Tech.eu’s summit coverage and Fortune Startups’ executive interviews independently echoed the same broad pattern: many firms run pilots, but few have integrated AI into core operations or reached mature production use.

What remains uncertain

The coverage does not provide the study’s methodology, sample composition, or sector breakdown, so it is unclear how broadly the 88% figure applies across industries and company sizes. Some supporting coverage is opinion-based or sponsored, and the articles do not quantify the cost, timeline, or organizational changes required to close the build-vs.-run gap.

Monitor next

Watch for publication of the underlying IDC-Lenovo methodology or follow-on benchmarks showing whether organizations that add governance, monitoring, and operational controls improve pilot-to-production conversion rates.

Analytical support, not advice — assumptions and open questions stated above.

Source coverage